How to Make Better Decisions: The Statistics Behind Everyday Choices
"87% of people agree!" "Studies show..." "Experts say..." We're bombarded with statistics every day, but most of us can't tell which ones are legit and which ones are complete garbage. Let's fix that.
You see a headline: "New study proves coffee causes cancer!" A week later: "Coffee prevents cancer, scientists say!" What the hell is going on?
Or you're choosing between health insurance plans. One costs $200/month with a $2,000 deductible. Another is $300/month with a $500 deductible. Which is actually cheaper? (Hint: most people get this wrong.)
Here's the problem: we make dozens of decisions every day based on data, but we never learned how to actually think about statistics. We can calculate a mean in Excel, sure. But can we spot when someone's using statistics to lie to us?
Let's fix that. This isn't about formulas or equations—it's about developing a BS detector for the modern world.
Red Flag #1: Tiny Sample Sizes
The Headline:
"Study finds people who eat chocolate daily live longer!"
The fine print: Study included 12 participants over 3 weeks.
12 people. That's barely a dinner party. Yet media outlets run with it because "CHOCOLATE MAKES YOU LIVE LONGER" gets clicks.
Why Sample Size Matters
Imagine flipping a coin 4 times. You get 3 heads, 1 tails. Does that mean the coin is biased? No—with small samples, randomness dominates.
Now flip it 1,000 times. If you get 750 heads, that's evidence of bias. Larger samples smooth out random fluctuations and reveal true patterns.
Rule of Thumb
n < 30
Too small—ignore it
30-100
Suggestive, not conclusive
100+
Now we're talking
Quick Check
Next time you see a study, ask: "How many people participated?" If the article doesn't mention it, that's a red flag in itself.
Red Flag #2: Correlation Doesn't Mean Causation
This is the #1 way statistics lie to you. Two things happen together? Must be related, right? Wrong.
Real Examples of Spurious Correlations
📈Ice cream sales correlate with drowning deaths. Does ice cream cause drowning? No—both increase in summer.
📊Nicolas Cage movies correlate with pool drownings. Yes, really. It's random coincidence with enough data.
📉Divorce rate in Maine correlates with margarine consumption. Seriously. Look it up.
The Three Possibilities When Two Things Correlate
Option 1: A causes B
Smoking causes lung cancer. High test scores cause college acceptance. This is what most headlines assume.
Option 2: B causes A
Maybe the causation goes the other direction. "Wealthy people live longer" could mean money causes health—or healthy people are more able to earn money.
Option 3: C causes both A and B
A hidden third variable is the real cause. Example: "Coffee drinkers have more heart attacks" sounds bad until you realize coffee drinkers also tend to be stressed, sleep-deprived workers.
Option 4: Pure coincidence
With enough data points, random patterns emerge. That Nicolas Cage thing? Just noise.
How to Spot It
Look for randomized controlled trials (RCTs). That's where researchers actually test causation by randomly assigning people to groups. If a study isn't randomized, it can only show correlation, not cause.
Red Flag #3: Ignoring Base Rates
This one is sneaky. It's how medical tests freak people out unnecessarily.
Real Scenario
You test positive for a rare disease. The test is 95% accurate. Should you panic?
Most people say: "Yes! 95% accurate means I probably have it!"
Statistics says: "Actually, you probably DON'T have it."
The Math That Saves Your Sanity
Let's say the disease affects 1 in 1,000 people (0.1%)
Out of 10,000 people:
• 10 actually have the disease (0.1%)
• The test correctly identifies 9-10 of them (95% accurate)
But here's the kicker:
• 9,990 people DON'T have the disease
• The test falsely flags 5% of them = ~500 false positives
Result: Out of ~510 positive tests, only 10 are real.
Your chances of actually having the disease? About 2%, not 95%.
This is called Bayes' Theorem, and it's why doctors order follow-up tests before diagnosing rare conditions.
The Lesson
Always ask: "How common is this thing in the first place?" A 99% accurate test for something incredibly rare will still produce mostly false positives.
Real-World Application: Choosing Insurance
Remember that insurance question from the intro? Let's solve it with expected value—the most useful statistical concept you never learned.
The Choice
Plan A
$200/month
$2,000 deductible
Plan B
$300/month
$500 deductible
Calculate Expected Value
If you're healthy (need minimal care):
Plan A: $200 × 12 = $2,400/year
Plan B: $300 × 12 = $3,600/year
Winner: Plan A saves you $1,200
If you have a major expense ($5,000 in medical bills):
Plan A: $2,400 premiums + $2,000 deductible = $4,400 total
Plan B: $3,600 premiums + $500 deductible = $4,100 total
Winner: Plan B saves you $300
The Decision:
If you're young and healthy, Plan A saves $1,200. If you expect high medical costs, Plan B saves $300. The break-even point? About $8,000 in annual medical expenses.
This is expected value thinking: weighing costs against probability. Insurance companies use it to set prices. Now you can use it to make better choices.
Red Flag #4: Cherry-Picked Data
"Our product has a 95% satisfaction rate!" Sounds great. But what if they surveyed 20 people and only counted the 19 who liked it?
Common Cherry-Picking Tactics
•Selective time frames: "Sales up 300% this month!" (ignoring the 90% drop last quarter)
•Ignoring contradictory data: Publishing only positive studies, burying negative ones
•P-hacking: Running 100 analyses until one shows significance by chance
•Survivorship bias: Only counting successes that made it to the end
Real Example: Startup Success Stories
You read about a founder who dropped out of college and built a billion-dollar company. The conclusion? "College is a waste of time!"
What they don't show you: the thousands of college dropouts whose startups failed and who now struggle without a degree. This is survivorship bias—we only see the winners.
How to Spot It
Ask: "What's the full picture?" and "Who didn't make it into this statistic?" Look for representative samples—data that includes failures, not just successes.
Real-World Application: Online Reviews
You're buying a product online. Two options:
Product A
Perfect 5.0 rating
Product B
4.3 rating
Most people pick Product A. "Perfect 5 stars!" But statistically, Product B is the safer bet.
Why?
- • Sample size: 15 reviews could easily be friends/family or cherry-picked
- • Regression to the mean: With more reviews, Product A's rating would likely drop
- • Confidence interval: We're much more confident in the 2,847-review average
A 4.3 rating with thousands of reviews tells you: "This product consistently satisfies most people." That's valuable information.
The Bottom Line: Think Like a Statistician
You don't need to be a math genius to spot statistical BS. Just ask these five questions:
1. How big was the sample?
2. Does correlation actually prove causation, or could there be other explanations?
3. What's the base rate (how common is this thing)?
4. What data are they NOT showing me?
5. Who's paying for this study, and what do they have to gain?
Statistical literacy isn't about memorizing formulas. It's about developing a healthy skepticism for claims that sound too good (or too scary) to be true.
The next time someone tells you "studies show," don't just nod along. Ask the questions above. Dig into the details. Think critically.
Because in a world drowning in data, the ability to separate signal from noise is a superpower.